Commit
•
5fdd23f
1
Parent(s):
f76d700
create README
Browse files
README.md
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- digitalpipelines/wizard_vicuna_70k_uncensored
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
# Overview
|
| 8 |
+
Fine-tuned [OpenLLaMA-7B](https://huggingface.co/openlm-research/open_llama_7b) with an uncensored/unfiltered Wizard-Vicuna conversation dataset [digitalpipelines/wizard_vicuna_70k_uncensored](https://huggingface.co/datasets/digitalpipelines/wizard_vicuna_70k_uncensored).
|
| 9 |
+
Used QLoRA for fine-tuning using the process outlined in https://georgesung.github.io/ai/qlora-ift/
|
| 10 |
+
|
| 11 |
+
# Prompt style
|
| 12 |
+
The model was trained with the following prompt style:
|
| 13 |
+
```
|
| 14 |
+
### HUMAN:
|
| 15 |
+
Hello
|
| 16 |
+
|
| 17 |
+
### RESPONSE:
|
| 18 |
+
Hi, how are you?
|
| 19 |
+
|
| 20 |
+
### HUMAN:
|
| 21 |
+
I'm fine.
|
| 22 |
+
|
| 23 |
+
### RESPONSE:
|
| 24 |
+
How can I help you?
|
| 25 |
+
...
|
| 26 |
+
```
|